What's screwed up about cpu speed comparison charts

You notice that he does also focus on 16bit hardware a lot too? And that maybe given the cost of 16bit systems that maybe people who ran them might like to know how quickly a system does add numbers together?

As for the method, I suggest you go an read up on the PDP-11 and VAX systems, but basically it's more comparing like with like so the results can't be skewed (much) one way or the other.

 

One thing I always wondered is what so great about 32-bit math that you couldn't do with 16 and 8-bit math? I mean seriously, do you really need to round screen coordinates to 32 decimal places?

I just don't get it?

 

32bit math makes for a nice 16bit:16bit fixed point addition (register to register part on the 68k original is the same as 16bit, but you still have some overhead of loading storing 32bit vs 16bit). Yeah, it might be a overkill, but there's no need for shifting when you only need to load the 'whole' part, unlike 12bit:4bit fixed point 16bit add. They're most likely not going to be onscreen coordinates, but you might have 'objects' that free roam, or similar, around in the virtual map. It's nice and it's quick.

But 32bit math is touted waaaaay too much. The majority of arithmetic could be done in 16bit or even 8bit no problem. No just arithmetic, but compares, bit testing/setting, conditional branching, etc. On the 68k it's common place though (32bit math/access). It's there and actually usable/used, instead of trying to find a way avoid using it and reducing operations to 16bit or 8bit. An example, 32bit math is common for building an address + base (because the other indexed addressing modes are less effective cycle wise) and then using auto inc/dec on the reg. On other processors, they have a free and efficient indexing system and have no such need for 32bit adding of offsets. You get use to one processor and tend to fall in the trap of thinking in terms of optimizations/style of that processor when evaluating other processors.


But, I don't to make this out to be a 68k bashing post/response. I actually really like coding on the 68k - it's different but a lot of fun. The only problem I have seems to come from the Amiga community (and a little from the ST community too). And their misplaced biased/ignorant view that the the 68k was king shit and nothing compared back in the day. Pretty arrogant attitudes.